Re: libstdc++ dependency mis-detection in gcc packages?

[Jonathan Perkin <jperkin%mnx.io@localhost>]

* On 2023-05-02 at 11:56 BST, Havard Eidnes wrote:

> 2) The main point: is the pkgin warning for the gcc12 and gcc10
>   package installation spurious?  Is the bug with the generation of
>   pkg_summary?  We really should not emit a warning which seems to
>   indicate "this will not work" when it evidently does anyway.
>   Teaching users to ignore warnings is bad practice.

As I'm writing this email I'm getting strong deja vu, I think this came 
up recently and may have been something to do with the packages being 
built inside a chroot with the wrong sets unpacked?  Anyway, worth 
checking the recent archives.

As to the first question, this is nothing to do with pkgin itself, or 
any other tool that uses pkg_summary, as they only use the information 
provided, so yes if there's a legitimate bug here it is in the 
generation of pkg_summary.

The generation of REQUIRES comes from the ${_BUILD_INFO_FILE} target in 
mk/pkgformat/pkg/metadata.mk.  On most systems it does a basic ldd on 
all of the binaries and libraries that are contained in the package, and 
so on systems where this is recursive it may end up listing libraries 
that are not directly required by the package (and so the software may 
work regardless).

For many years I've used this patch which instead only records direct 
dependencies via DT_NEEDED:

  https://github.com/TritonDataCenter/pkgsrc/commit/dff5fd726aa1d33966ab4557cc55026f38905f46

It might be good to get consensus on this and figure out a portable way 
to do it for all ELF platforms.

--
Jonathan Perkin   -   mnx.io   -   pkgsrc.smartos.org
Open Source Complete Cloud   www.tritondatacenter.com